How a lunch resulted in an AI-Powered Solution to support clean water access in Uganda. A guest post written by Simon Kloker PhD.
A sunny day in spring 2024 in Uganda. Well, April or March. There is no real “spring” in Uganda and basically almost every day is sunny. However, let it just be a day as any. I'd lived in Uganda for almost a year at this time, and I'd grown accustomed to the local customs, unwritten rules, and yes, even the daily struggle for food. I work at Ndejje University as a Consultant for Artificial Intelligence (AI), employed by a German Christian organization – Coworkers – for International Development Cooperation. Once or twice a week, I would use the university's staff cafeteria, where local food is free. There is just one thing to know and one rule to follow – you should know that you never know when food will finally arrive. Prepare yourself for waiting, and make it your habit as soon as you see a vehicle approaching, to secure your place in the queue. Food will likely not be enough when all those waiting want their share – and if you don’t want to just skip the line as many locals do – you need to have at least a place in the first third, to end up among the last that gets served at least something.
This particular day, there was another Muzungu in the cafeteria. “Muzungu” refers to a white and surely very rich guy, for through Ugandan eyes every white guy must be very rich. He was surrounded by a talkative group of Ugandans. It is a rare occasion, but since he did not seem lost, I just continued my own routine. When the food finally arrived and I got my top-position in the queue, I felt a bit sorry for this poor guy who ended up at the very end of the queue. Maybe Muzungu-Bonus will save him and they will keep something back. Actually, I never asked him whether he got something to eat that day 😊. Anyway, I took my place and ate my lunch. Quarter an hour later, Joshua, as I learned was his name, approached my table – to greet the “other Muzungu” in the room. We exchanged some basic pleasantries, shared contact information, and agreed to meet again.
We met in my office, at a later date, and had Rolex for lunch (just google it). At our lunch, I learned about his work and realized we shared some fundamental principles, values, and experiences. It is always a gift meeting someone to whom you can share your successes, struggles and all those stories that need to be told, but can only be understood by someone in a similar situation. Such people are rare, especially when you work as a westerner in the middle of Africa. I learned that he founded an organization focusing on clean water access in Uganda – The Ayin Project – and also learned some of his "stories" that reveal so much about what truly drives their work and vision. Joshua was very eager for new technologies like AI, and I'm always looking for use cases relevant to local communities. I agreed to brainstorm ideas on how to use AI to improve clean water access in Uganda.
The problem with AI in Uganda is the lack of data. There are no reliable data sources, registries, protocols, or sensors. Only a few things are digitized. This is the same problem he faces in his work. It's difficult for him to know about the state of knowledge in local communities on WASH or the condition of their wells and boreholes, especially in remote areas. These places are scattered and cannot be reached easily. Whether you want to learn about the current condition or provide knowledge and education, a scalable solution can't rely on visiting each location. When the semester ended in June, I had some extra time and had just finished another project. I considered looking for final-year students interested in working on water-related thesis topics, but finally failed. However, the idea to use AI in connecting with people suffering from water issues persisted.
Smartphones are becoming increasingly common, even in remote villages that cannot be reached without a 4-wheel drive vehicle. Likewise, WhatsApp is also widely available. Though it’s surely not omnipresent as in most of the northern hemisphere, but in Uganda, in each extended family you will somehow find one smartphone. I decided to take a new project upon myself. I arrested two dear local colleagues to work with me on the idea to reach these people through WhatsApp and AI. Through this WASHtsApp was born. WASHtsApp is easy – you just start chatting via WhatsApp to a Ugandan phone number and it can respond and suggest questions regarding WASH issues. The chatbot is powered using an approach called Retrieval Augmented Generation. Just think of ChatGPT or Google Gemini, with the difference that you force the AI to only provide responses based on a context that you provide. This context was WASH-related information for rural areas in developing countries. It’s that easy.
We evaluated WASHtsApp's response quality and user acceptance through a small-scale study and survey-based interviews with people at public wells. The feedback was overwhelming. The last question in the survey asked: “Anything you want to let us know?”. We were flooded with answers like:
"My area is in dire need of safe water and sanitation facilities" There is a lot of water stress in my area. "No water reservoir in the whole district." In Bugiri, water is a problem. "We have insufficient water supply." Lack of water. "Society doesn't have enough information about clean water." We need more water. "Come to communities." We are suffering.
“We are suffering”. Sometimes we sit in our offices on comfortable chairs (even if I have to note that mine here in Uganda is not too comfortable 😊), thinking of AI and whatever else – while people are quietly suffering from the lack of very basic needs. However, there was one answer to this question mentioned even more often: Sincere gratitude for our care for their situation.
The project is still in its early stages, but Joshua and I are committed to making this solution truly beneficial to local communities on a large scale. If there is a seed of hope – provided with clean water – it can grow.
Want to learn more? We submitted the study to IST Africa Conference 2025 and hope to present it there. A working paper version is accessible here: https://arxiv.org/abs/2411.02850
Comments